【作者】 刘艳；

【导师】 潘善亮；

【作者基本信息】 宁波大学 ， 计算机应用技术， 2015， 硕士

【摘要】 面向服务的计算发展日新月异,面对互联网海量的功能属性相近的服务,如何从大量服务中选择出符合用户自身个性化偏好的服务,如何从候选服务中为使用者推荐迎合其需求的服务,这成为了Web服务推荐的关键技术。当前的社交应用平台如facebook、推特、新浪微博以及基于位置的社会化网络(Location-Based Social Networks,LBSN)等已普遍应用,用户数量日益增多,社交网络的使用者如何快速检索到符合自己兴趣偏好的服务信息已成为一个备受关注的话题。传统的个性化推荐方法,如协同过滤方法等,虽有不俗的表现,然因社交网络自身的特点,这些传统的方法的准确度大打折扣。本文关注社交网络中用户关联、用户兴趣传播、好友之间信任传递以及项目本身属性对推荐结果的影响,提出一种利用LBSN数据和使用者关系来产生推荐列表的方式,利用社交环境中用户们上传的景点照片等信息评估用户偏好;利用好友亲密度、用户景点评分寻找其相似邻居;在考虑到使用者兴趣偏好的同时也考虑景区的适宜旅行时间以及候选景区附近的其他地区的推荐。现有使用社会性网络服务(Social Networking Services,SNS)的信任关系推荐时通常仅考虑到用户社会关系或服务全局评分中的某一方面。实际中用户在选择各式各样的服务时,往往选择信誉良好、综合评分高的服务,也会兼顾信任好友的意见。本文探索如何在海量功能相似、评分差别不大的服务中利用信任关系和兴趣偏好获得推荐列表。当目标使用者和好友都未选择过目标服务时,选用服务的流行度即全局评分以及兴趣偏好作为推荐依据;而当好友选择使用过服务时,利用服务的全局评分和用户所信赖好友的意见即本地信任评分,按照合理的策略组合预测形成最终符合个性化要求的推荐列表。最后通过实验,对所提出方法进行有效性验证。更多还原

【Abstract】 With the continuous development of service-oriented computing, there are always constantly emerging new service that having same functional properties with each other. The key technology of Web Services Recommendation is how to rapidly choose services from the candidates that meet the demands for target users.The current social networks such as Facebook, Twitter and location-based social networking(LBSN), have been widely used with the increasing number of users. How to quickly retrieve service information in line with their own interests and preferences has become a topic of concern for users. Despite the traditional method of personalized recommendation has a good performance, however, due to the social network’s own characteristics, the accuracy of these traditional methods are greatly reduced. This article focuses on the association of social network users, the interest transfer,and the affection of the attributes of project itself. This paper proposed a method that use data and users’ relations between LBSN to generate a list of recommended, using users’ photos of scenic spot to assess their preferences; using intimacy and user rating data for scenic spots to meet the similarity uses. Taking into account the interest of the users’ preferences with the suitable travel time of recommended location as well as other suitable areas nearby the candidates when making recommendation.Existing studies using SNS trust usually only consider the relationship between the user community or service in a particular aspect of the overall score. These methods had some drawbacks. When selecting services from a vast amount of services, users often choose services with high composite score and good reputation meanwhile, users will also refer their friends’ comments. This article was explored that how to choose services from lots of services with similar functions to meet the requirements is a question. We have to take into account the popularity of the global score when the target user and his friends had not selected a target service. However when a user used to select the service, we also adopt the local trust with his friends’ comments, then combined the strategy to form the recommended list. Finally, we used experiments to verify the validity of the proposed methods.更多还原